TL;DR: Terafab is a $20–25 billion chip manufacturing project jointly launched by Tesla. SpaceX and xAI on March 21, 2026. It aims to produce up to 200 billion AI chips per year with 80% of that compute destined for a constellation of up to one million orbital AI data center satellites in low Earth orbit.
What Is the Terafab Project?
It is a $20–25 billion semiconductor fabrication facility announced by Elon Musk on March 21, 2026. In Austin Texas. It is a joint venture between Tesla. SpaceX and xAI and is designed to be the largest chip manufacturing facility ever built.
Unlike traditional chip projects that rely on third-party manufacturers like TSMC or Samsung. Terafab will handle the entire chip production pipeline under one roof from design and lithography to fabrication, memory, advanced packaging and testing. The facility will produce 2-nanometer chips the most advanced node in current commercial production at a projected rate of 100 to 200 billion chips per year.
Musk’s driving reason for building it? Existing chip suppliers can only meet 2% of what Tesla and SpaceX need to execute their AI ambitions. So rather than wait, he decided to build the supply chain himself.
What Are the Two Chips Terafab Will Produce?
It is engineering two distinct chips for two very different environments:
1. The AI5 Chip — Built for Earth
The AI5 is Tesla’s next-generation inference chip, purpose-built to power:
- Full Self-Driving (FSD) software
- The Cybercab robotaxi fleet
- Optimus humanoid robots
Compared to the current AI4 chip, the AI5 delivers 40–50x more compute performance and 9x more memory bandwidth. Small-batch production is targeted for late 2026, scaling to full volume through 2027.
2. The D3 Chip — Built for Space
The D3 is engineered specifically for orbital environments, designed to withstand:
- High-energy radiation and particle bombardment
- Extreme temperature fluctuations
- The harsh vacuum conditions of low Earth orbit
It forms the foundation for Musk’s most ambitious vision yet. A space-based AI computing network that makes traditional ground data centers look like dial-up internet.
Why Is 80% of Terafab’s Compute Going to Space?
This is the part most people aren’t talking about enough. Musk doesn’t just want to build better chips on Earth. He wants to relocate most of humanity’s AI computing to low Earth orbit.
Here’s why he believes space wins economically within 2 to 3 years:
- Solar energy is ~5x stronger in space than on Earth’s surface and it never stops — no night cycles, no clouds, no seasonal drops
- Cooling is virtually free — heat dissipates naturally in the vacuum of space, eliminating one of the largest operating costs of ground-based data centers
- Starship launch costs are collapsing — SpaceX’s reusable rockets are making it increasingly viable to put hardware in orbit at massive scale
The math, according to Musk, eventually makes space cheaper than Earth for running AI workloads at scale.
What Are the AI Sat Mini Satellites?
SpaceX’s orbital AI satellites are called AI Sat Mini. Each one is:
- Over 170 meters long — roughly the length of two Boeing 747s end to end
- Powered by massive solar arrays generating 100 kilowatts each with future versions scaling to megawatt capacity
- Designed to function as fully autonomous orbital data centers
SpaceX has already filed with the FCC to deploy up to one million of these satellites into low Earth orbit. Combined, they would form the largest distributed computing network ever created. floating silently above Earth, processing AI workloads around the clock.
What Is Elon Musk’s Long-Term Vision for Terafab?
Musk described Terafab as the next step toward becoming a galactic civilization. The long-term roadmap includes:
- Scaling to 100 million tons of solar capture equipment sent into space annually
- Extending computing infrastructure beyond Earth orbit, with early discussions around lunar data centers
- Preserving human memories and knowledge across generations through AI. a concept Musk framed as a path toward digital immortality
Whether you find that inspiring or unsettling likely depends on how much you trust his timelines. But the infrastructure groundwork site preparation, FCC filings and early hiring is already visibly underway.
Should You Be Skeptical About Terafab?
Healthy skepticism is fair here. Musk has a well-documented history of ambitious timelines that slip. Building a world-class semiconductor fab from scratch. Without existing expertise in chip manufacturing is an extraordinarily difficult undertaking. The AI5 chip itself was delayed even before Terafab was announced.
But unlike some past announcements, this one comes with concrete early-stage activity:
- Drone footage confirms large-scale site preparation at Giga Texas North Campus
- Tesla has begun posting Terafab-specific job listings
- SpaceX’s FCC application for the satellite constellation is formally filed
- Musk himself put it plainly: Either we build Terafab, or we can’t get the chips — so we build the Terafab.
Terafab Fast Facts
| Detail | Figure |
|---|---|
| Total Investment | $20–25 billion |
| Location | Austin, Texas (Giga Texas North Campus) |
| Companies Involved | Tesla, SpaceX, xAI |
| Annual Chip Output | 100–200 billion chips |
| Chip Process Node | 2 nanometers |
| Total Compute Target | 1 terawatt (1 trillion watts) |
| Space vs. Earth Split | 80% space / 20% Earth |
| Orbital Satellites Planned | Up to 1 million |
| Power per Satellite | 100 kW (solar) |
| Production Start | Late 2026 (small batch), 2027 (volume) |
Frequently Asked Questions About Terafab
What does Terafab stand for?
It is short for Tera Fabrication. A reference to its goal of producing a terawatt (one trillion watts) of computing power in chips annually.
Who owns Terafab?
Terafab is a joint project between Tesla, SpaceX and xAI all companies closely associated with Elon Musk.
Where will Terafab be built?
It will be built in Austin, Texas, starting at the historic Seaholm Power Plant site and expanding to the North Campus of Gigafactory Texas.
How does Terafab affect NVIDIA and TSMC?
If successful, Terafab would significantly reduce Tesla and SpaceX’s dependence on both TSMC for chip manufacturing and NVIDIA for AI accelerators. Representing a major vertical integration move in the semiconductor industry.
When will Terafab chips be available?
Small-batch production of the AI5 chip is targeted for late 2026. with full volume production expected in 2027.
What is the D3 chip used for?
The D3 chip is a radiation-hardened processor designed specifically for SpaceX’s orbital AI data center satellites. Intended to operate reliably in the harsh conditions of low Earth orbit.
Whether Terafab becomes the defining tech infrastructure story of the next decade or another chapter in Musk’s long book of ambitious timelines. one thing is clear the AI chip wars just entered a new dimension. And the battleground isn’t just Silicon Valley anymore. It’s low Earth orbit.
Follow TechGlimmer.io for ongoing coverage as Terafab develops.
Sources
- Electrek — Tesla and SpaceX announce $25B ‘Terafab’ chip factory
- Tom’s Hardware — Elon Musk formally launches $20B Terafab chip project
- The Verge — Musk says he’s building Terafab chip plant in Austin, Texas
You might be interested in following article
What Is a Quantum Battery? CSIRO Just Built the World’s First One
